/JTFT

A Joint Time-frequency Domain Transformer for Multivariate Time Series Forecasting

Primary LanguagePythonApache License 2.0Apache-2.0

JTFT

This is an implementation of JTFT: "A Joint Time-frequency Domain Transformer for Multivariate Time Series Forecasting."

Usage

  1. Install requirements. pip install -r requirements.txt

  2. Download data. You can download all the datasets execept for PEMS from Autoformer. Create a seperate folder ./dataset and put all data files in the directory. The PEMS data can be downloaded from https://github.com/zezhishao/BasicTS. The npz files in the dataset can be processed to csv files using data_proc_pems.ipynb.

  3. Training. All the scripts are in the directory ./scripts/. The scripts can be run using commands such as

sh ./scripts/weather.sh

The results will be displayed in the log files once the training is completed. The path of the log files will be printed at the beginning of the training.

Acknowledgement

We appreciate the following github repo very much for the valuable code base and datasets:

https://github.com/yuqinie98/PatchTST

https://github.com/cure-lab/LTSF-Linear

https://github.com/zhouhaoyi/Informer2020

https://github.com/thuml/Autoformer

https://github.com/MAZiqing/FEDformer

https://github.com/alipay/Pyraformer

https://github.com/ts-kim/RevIN

https://github.com/timeseriesAI/tsai

https://github.com/zezhishao/BasicTS

License

Some of the codes are obtained from https://github.com/yuqinie98/PatchTST. These files are licensed under the Apache License Version 2.0.

The new files of JTFT are licensed under the GNU General Public License (GPL) version 2.0. Comments to show the license appears at the beginning of these file.